Exploring the Fusion of Graph Theory and Diverse Machine Learning Models in Evaluating Cybersecurity Risk

The frequency and severity of cyber- attacks have surged, causing detrimental impacts

on businesses and their operations. To counter the

ever-evolving cyber threats, there's a growing need

for robust risk assessment systems capable of

ef ectively pinpointing and mitigating potential

vulnerabilities. This paper introduces an innovative

risk assessment technique rooted in both Machine

Learning and graph theory, which of ers a method

to evaluate and foresee companies' susceptibility to

cybersecurity threats. In pursuit of this objective, four Machine Learning algorithms (Random Forest, AdaBoost, XGBoost, Multi-Layer Perceptron

(MLP)) will be employed, trained, and assessed

using the UNSW-NB15 dataset that has a hybrid of

real modern normal activities and synthetic

contemporary attack behaviours..The findings

indicate that the Multilayer Perceptron (MLP)

performs better than other classifiers, achieving an

accuracy of 98.2%.. By harnessing the capabilities

of data-derived insights and intricate network

analysis, this groundbreaking approach aims to

equip organizations with a comprehensive and

forward-looking cybersecurity defense strategy.

Read Download